EU says AI Act can handle rogue model risks

EU tech chief Henna Virkkunen said the bloc is “well equipped” to address risks from rogue AI systems under the AI Act, which was adopted two years ago. Concerns have grown after recent incidents at OpenAI and Anthropic raised fears that highly capable systems could slip beyond human control.

The AI Act regulates technology according to risk and covers the full life cycle of advanced models, Virkkunen said. Regulators are preparing guidance for companies on model evaluation and timing, while also considering recommendations from a scientific panel of 60 AI experts from universities and institutions.

The European Commission sent requests for information in late August to more than 30 AI companies, seeking details on safety and security measures before following up on transparency and copyright obligations. Those requests can lead to investigations and fines of as much as 7% of a company’s global annual turnover.

Virkkunen said Chinese AI startups also received information requests, noting that the most capable models now on the market come from the United States and China. A possible digital levy aimed largely at big US tech companies remains in the hands of EU governments, which are discussing financing sources for the EU budget for 2028 to 2034.